Gene cloning to clinical trials—the trials and tribulations of a life with collagen

Summary

This review, based on the BSMB Fell‐Muir Lecture I presented in July 2018 at the Matrix Biology Europe Conference in Manchester, gives a personal perspective of my own laboratory's contributions to research into type X collagen, metaphyseal chondrodysplasia type Schmid and potential treatments for this disorder that are currently entering clinical trial. I have tried to set the advances made in the context of the scientific technologies available at the time and how these have changed over the more than three decades of this research.

1. INTRODUCTION

Type X collagen is a collagen with an interrupted triple helical domain that is approximately half the length of the fibrillar collagens, and it is synthesized almost exclusively by hypertrophic chondrocytes in endochondral growth plates. In this review, I will address key aspects of the research I was involved with relating to type X collagen and I will explain the difficult decisions we had to grapple with in the context of the research field at that time. Many of the achievements described may be considered trivial by today's standards, but the reader should bear in mind that when the narrative begins (1981): human genetic screens were conducted by restriction fragment length polymorphism (RFLP) analyses on Southern blots; PCR would not be invented for another 5 years; DNA sequencing was performed manually generating 300 base reads per run; Sanger had recently won the Nobel prize for dideoxy DNA sequencing; the Human Genome Project did not exist; there was no Internet; and there were no desktop computers or mobile (or smart) phones or email. Furthermore, researchers spent hours going to the library to do literature searches and read and photocopy journal articles that were of interest.

2. THE EARLY YEARS: 1981‐1989

Type X collagen was discovered independently in the early 1980s by two laboratories at The University of Manchester, UK,1, 2, 3 and The University of Illinois, USA,4, 5 respectively. It was about this time (1981) that, fresh from my doctoral studies in London, I joined the extracellular matrix research group in Manchester led by Mike Grant. My postdoc project was to learn about basement membrane synthesis and continue my doctoral research into the aetiology of diabetic microangiopathy. Whilst the work on type X collagen in the laboratory was making great headway with Cay Kielty as the project postdoc, my own research was slow‐going and it gradually became clear that we needed the latest “molecular biology” approaches to make significant headway. A serendipitous meeting of my laboratory boss (Mike Grant) with Darwin Prockop in 1984 led swiftly to me resigning my postdoc in Manchester and starting 1985 in the USA at Rutgers Medical School in Piscataway NJ. My project was to utilize newly cloned cDNAs coding for type IV collagen and laminin to measure mRNA levels in tissue extracts to determine the role of changes in basement membrane synthesis in the aetiology of diabetic microangiopathy. Whilst nowadays one would do RNASeq and analyse the entire transcriptome using systems biology techniques,6, 7 in 1985 this was cutting edge science and I obtained a fellowship from the American Diabetes Association and other funding from charities and ICI (now AstraZeneca) to support the work—my independent career had started. The work at Rutgers and subsequently in Philadelphia, following the move of the whole Prockop group in April 1986 to form The Jefferson Institute for Molecular Medicine, went well resulting in several papers.8, 9, 10 During this time, I also learnt how to screen cDNA libraries—cloning genes and determining their sequence was an important research activity in the eighties and would continue to be an important skill to have in your molecular biology repertoire until the mid‐nineties when genome sequences started to become available on the rapidly expanding Internet. I had made sufficient progress during my 3‐year stay in the Prockop Lab to convince the Royal National Institute for the Blind to fund me on a Fellowship to return to Manchester and set up my own laboratory within the Matrix Group led by Mike Grant.

By 1987, many of the chick cartilage collagen genes had been cloned by the Olsen Lab in Harvard and the focus was turning to the mammalian/human genes since these sequences were the prerequisite for characterizing and diagnosing heritable diseases of man affecting connective tissues. Whilst some of the human cartilage collagen genes had been cloned recently,11, 12, 13 type X collagen had proved elusive. Upon my return to Manchester in late 1987, it was apparent that the work in Mike's Lab on collagen X was poised to make major advances with the application of the gene cloning techniques. Terrig Thomas was finishing off his PhD on cell biological aspects of chick type X collagen and keen to learn molecular biology techniques.14 Anne Marriot working with Shirley Ayad had developed a foetal bovine growth plate culture system that synthesized bovine collagen X,15 and Alvin Kwan had developed an antibody to purified chick type X collagen that cross‐reacted with bovine collagen X.16 Mike and I wrote a collaborative grant to the Nuffield Foundation (Oliver Bird Fund) to clone and characterize bovine collagen X cDNAs which was funded. Terrig was shown by Anne how to grow the foetal bovine growth plate chondrocyte cultures. Once they looked “ready,” Terrig would label the cells with radioactive proline for 24 h, collect the medium and isolate the total RNA. If, when analysed by SDS‐PAGE fluorography, the medium had collagen X present, then the RNA from that culture was added to the stock RNA that we knew “contained the mRNA for bovine collagen X.” The RNA would then be reverse‐transcribed to cDNA with reverse transcriptase and cloned into a λgt11 expression vector that could be immune‐screened using Alvin's collagen X antibody. The collection of RNA went smoothly, but the first major hurdle occurred when we were ready to have the cDNA library prepared. I had (naively) assumed that making the cDNA and preparing the library using newly developed commercial kits would be simple and we had accordingly budgeted very little in the consumables for this purpose. However, in the interim time after submitting the grant, I had used up several kits trying to make my own cDNA libraries with absolutely no success and wasting RNA in the process. The alternative was to go down the commercial route and send the RNA off to Clontech who were about the only biotech company at that time with an established and reliable reputation for making high‐quality cDNA libraries. The problem was that the cost of the commercial synthesis would virtually wipe the consumables budget for the entire project. Nevertheless, on talking the situation through with Mike Grant, he said “spend the money and get the library made.” We followed this sound advice, and two months later, the bovine growth plate λgt11 library was delivered. The first round of screening produced so many immune‐positive clones that at first, I was doubtful whether this could be real. However, Terrig purified several clones to homogeneity, subcloned the inserts into a sequencing vector and ran the first sequencing gels. It was immediately clear that the clones encoded collagens due to the repeating pattern of two Gs every 9 bases indicative of the presence of a glycine (whose codon is GGX) every third amino acid residue in the protein—the hallmark of a collagen triple helix. He took the X‐ray image of the sequencing gel home that evening, read the sequences, translated them and then compared the predicted protein sequences by eye with the previously published chick collagen X sequences.17 The high degree of identity between the bovine and chick protein sequences was immediately apparent, but the large amount of third base wobble in the mRNA sequences was clearly the factor that prevented cross‐species hybridization and precluded the cloning of the mammalian gene by simple screening using the chick cDNA probes. So, by Christmas 1988, just four months from starting the project, we had the first mammalian cDNA sequences for collagen X and the complete bovine sequence was characterized over the next six months with the help of a significantly increased consumables budget provided by The Nuffield Foundation.

Bovine and human gene sequences are far more closely related to each other than either is to chick, and therefore, the bovine cDNAs for collagen X could be used to screen a genomic library for the human gene. Normally, collagen genes such as those encoding the fibrillar collagens have relatively short exons (eg 54 or 99 bases) separated by relatively large introns and the whole gene can be dispersed over many kb of genomic DNA. Hence, most genomic libraries were produced in cosmid vectors that could accommodate approximately 40 kb of sequence per clone and therefore give you a reasonable chance of cloning a significant part of any gene in a single clone. An intriguing feature of the chick collagen gene was that almost all of the protein coding sequence for the gene together with the 3'untranslated region was encoded in a single long exon17—the gene looked like a cDNA sitting in the genome. We showed by Southern blotting that the bovine gene, like that of chick, was condensed18 and therefore, by implication, the human gene would most likely have the same conserved structure. We could therefore predict that the part of the gene coding for almost all of the human collagen X protein sequence may be contained in a 3‐kb segment of human genomic DNA. With the aid of a small grant from MRC, we therefore used the bovine collagen X cDNA to screen a human genomic library cloned into a λ vector which could accommodate approximately 10 kb of genomic sequence per clone. λ libraries were far easier to screen and subsequently analyse than cosmids because of their smaller genomic inserts. The screens were successful, and in a matter of weeks, we cloned and purified the 2 λ clones that subsequently yielded the full translated sequence of human COL10A1 gene.19

This was a very exciting time in the laboratory where almost everything we tried worked and we knew we led the field. On a personal level, my appointment to a lectureship in Manchester marked the end of “the early years” in my career and I certainly had no idea that this also marked the beginning of what turned out to be a 30‐year association with type X collagen and the growth plate.

3. THE MIDDLE YEARS (1990‐1999)

On a professional level, we were faced with a difficult ethical dilemma, namely, when to publish these sequences—in particular, that of the human COL10A1? At this time, it was important to be the first to publish a gene sequence—many a postdoc's career was cut short by not being able to publish their sequences because they were scooped by another group. We cloned the human gene with the intention of setting up a programme to identify human disease(s) caused by mutations in collagen X. Our dilemma related to the unusual condensed nature of the gene. PCR had been described three years previously and was now a routine technique in many laboratories. If we published the human sequence, anyone with a PCR machine could amplify up the complete COL10A1 gene from a genomic DNA sample and therefore conduct a genetic screen for mutations. This was not the case for almost all other collagen genes which had numerous exons and for which you also needed the intron sequences to effectively screen for mutations. Our future funding depended on us having exclusive use of the COL10A1 sequences at least in the short term.

So began a very tense two‐year period during which Mike Grant and I secured programme grant support from the Arthritis Research Council UK (now Versus Arthritis) for our next five years of research which included a genetic screen for diseases. We were fortunate to recruit Gillian Wallis to the programme to organize and conduct the genetic screen as she had extensive previous experience of screening for mutations in COL1A1 and COL1A2 associated with osteogenesis imperfecta. The screening strategy devised involved obtaining genomic DNA samples from groups of patients with candidate diseases (different forms of chondrodysplasia) through their consulting clinicians; amplifying the COL10A1 gene in five overlapping fragments; and then using DNA conformational techniques such as SSCP (single‐stranded conformational polymorphism) to determine whether the amplified fragments contained any base changes. These techniques were long‐winded and tedious, but they worked and were the most advanced available at that time. Large‐scale sequencing of these fragments was not possible since automated DNA sequencing using fluorescent tags was still in its infancy and, at this time, far too expensive to contemplate. Once a polymorphism was detected in a fragment by SSCP, the two alleles in the amplified sample had to be subcloned and sequenced manually. However, detecting a sequence change was not the guarantee of an interesting outcome. Many such changes affected the third base of codons and did not change the amino acid sequence. Others produced conservative substitutions of amino acid residues that were unlikely to be pathogenic. Once candidate pathogenic changes were identified, these needed to be shown to track with the disease in other family members (if other members were available). Alternatively, you needed to show that several unrelated individuals with the same disease had the same or similar changes in the gene that were not seen in any unaffected individuals. We eventually published the bovine cDNA sequences in early 199118 and were forced to submit the full‐length human COL10A1 sequences for publication19 when we learnt that Suneel Apte, then working in Bjorn Olsen's laboratory, had a short sequence of the human gene submitted for publication.20 The release of the human COL10A1 sequences enabled Gillian Wallis to publish two years of work detailing a number of chondrodysplasias that were not associated with COL10A1.21 The search for the disease caused by mutations in collagen X culminated when Mat Warman in Bjorn Olsen's group analysed a large Mormon kindred with metaphyseal chondrodysplasia type Schmid (MCDS) and found that every affected individual had a 13‐bp deletion in the region of the COL10A1 gene encoding the C‐terminal non‐collagenous domain that was not present in any unaffected family member.22 At the very moment this finding was announced by Bjorn Olsen at the Collagen GRC that year, Gillian Wallis in Manchester was subcloning fragments from two kindreds with Schmid that had given positive SSCP results indicating base changes. These cases turned out to represent pathogenic single amino acid substitutions, again localized to the C‐terminal non‐collagenous domain of the protein.23 Further analyses from our laboratory24 and many others, reviewed by Bateman et al,25 confirmed that MCDS is caused by COL10A1 mutations, mostly clustered in the C‐terminal non‐collagenous domain.

The description of MCDS‐causing mutations in the COL10A1 gene heralded the start of more than a decade of research searching for the mechanism by which these mutations caused the disease. Much of this research was conducted in either cell‐free translation systems or in cell culture. MCDS is a genetically dominant disease—that is, one mutant allele is enough to cause the disease. Some advanced the hypothesis that haploinsufficiency was the most likely mechanism based on the observations that mutant collagen X chains were very poorly secreted from cells and did not associate with wild‐type chains.26, 27 In addition, some mutations resulted in premature stop codons, and in at least two cases, these mutant mRNAs were unstable and destroyed by a chondrocyte‐specific mechanism related to nonsense‐mediated mRNA decay.28, 29 Conversely, the Col10a1 ko mouse had a very subtle phenotype that certainly did not greatly resemble MCDS in humans.30, 31, 32 Furthermore, stop codons anywhere in the COL10A1 gene would cause haploinsufficiency and yet none outside of the C‐terminal non‐collagenous domain have been described in MCDS patients. In fact, stop mutations in the region of the canine COL10A1 gene encoding the collagenous domain have been described more recently and shown to have no phenotypic effects.33 In addition, cell‐free translation showed that mutant collagen X peptide chains could trimerize with both wild‐type and other mutant chains.34, 35 These latter results argue for a dominant gain‐of‐function disease mechanism rather than haploinsufficiency. We also predicted the structure of the collagen X C‐terminal non‐collagenous “trimerization” domain34 based on the X‐ray crystallographic structure of the closely related ACRP30. With this structure, we were able to visualize the positions of amino acid residues mutated in MCDS and which were clustered in different subdomains of the trimeric C‐terminal domain rather that scattered randomly, again supporting a dominant gain‐of‐function rather than dominant negative‐type effect.34

4. THE ESCAPE (1999) AND LATER YEARS (2000‐2018)

Whilst much progress can be made using in vitro and cell culture techniques, the only way to be certain of how these disease‐causing mutations affect bone growth is to confirm their effects in vivo. Gene targeting to produce knock‐in models of disease‐causing mutation in mice by homologous recombination in ES cells was well established but by no means routine in the late 1990s. I was fortunate to have met Reinhard Fassler who was a leading exponent of these techniques in the extracellular matrix field and he kindly agreed to host me in his newly formed Department of Experimental Medicine in Lund, Sweden, to learn these techniques. With support from The Wellcome Trust and ARUK, I took a sabbatical and spent the whole of 1999 in Lund learning ES cell culture and gene targeting techniques. My wife and children accompanied me, and we all had a great experience living in Sweden for the year. Professionally, the sabbatical in Sweden felt like an escape from the teaching and administrative duties I had accumulated over the past decade in the Department in Manchester and enabled me to reacquaint myself full‐time with the bench.

On my return to Manchester in the new millennium, I began a collaboration with Mike Briggs who, several years previously, had joined what by now had evolved into the Wellcome Trust Centre for Cell‐Matrix Research. Mike Briggs's research was focussed on characterizing the disease mechanisms associated with mutations in COMP, matrilin‐3 and collagen IX that caused the multiple epiphyseal dysplasia/pseudoachondroplasia (MED/PSACH) spectrum of chondrodysplasias. He also wanted to investigate mouse knock‐in models and led a successful Wellcome Centre‐based programme grant application to NIH together with myself, Dave Thornton and Karl Kadler to make and phenotype a range of mouse models for MED/PSACH and MCDS. The NIH grant subsequently led on to Mike Briggs writing two successive multinational 5‐year EU grants (Eurogrow and SYBIL), which funded the major part of research in my laboratory for the remainder of my career.

In selecting the first mutations to target in mice for each disease, we decided that since many COMP/and matrilin‐3 mutations causing PSACH/MED were known result in the mutant protein being retained in the ER,36, 37 we would pick a MCDS‐causing mutation in the Col10a1 gene that might allow the protein to fold, be secreted and exert its pathogenic effects outside of the cell. Accordingly, we selected the N617K mutation in the Col10a1 gene since the asparagine residue in the wild‐type protein was predicted to be on the exposed surface of the assembled trimeric C‐terminal domain34 and the same residue in the closely related ACRP30 was indeed a lysine.

Hence, we and others38 were confident that the mutant N617K collagen X would fold correctly, trimerize effectively and be secreted. How wrong could we be in this assumption? The first indication that the N617K mutant collagen X may not be secreted came with the observations of Wilson et al that all of the MCDS mutants they tested in their cell culture system, including N617K misfolded, formed aberrantly disulphide‐bonded dimers and triggered the unfolded protein response (UPR) due to increased endoplasmic reticulum (ER) stress.39 Subsequently, Kathy Cheah and Danny Chan published papers on their transgenic MCDS mouse lines both of which showed retention of different mutant collagen Xs (13 base deletion and p.P620fsX621) triggering the UPR.40, 41 Indeed, their very detailed description of the growth plate pathology led them to hypothesize that the hypertrophic chondrocytes expressing mutant collagen X reversed their differentiation status to prevent collagen X expression as a mechanism to survive the induced ER stress.40 Likewise, our own studies characterizing the phenotype of the N617K MCDS mouse showed intracellular retention of the mutant protein resulting in a robust UPR triggered by increased ER stress.42 All of the mouse models shared the pathological features of an expanded hypertrophic zone in the growth plate and significantly reduced rates of long bone growth. Mechanistically, we went further and targeted increased ER stress and a resulting UPR to hypertrophic chondrocytes by generating transgenic mice expressing the cog mutant form of thyroglobulin from a collagen X promoter. Not only did these mice have the expanded hypertrophic zone associated with MCDS but also their bone growth rates were also reduced.42 We were now certain that the MCDS phenotype was driven directly by the increased ER stress and not by the absence of the wild‐type protein raising the possibility that reducing ER stress may be a way of treating MCDS and possibly other heritable diseases caused by ER stress‐inducing mutations in genes encoding components of the extracellular matrix.43, 44

Our research started to focus on methods for reducing ER stress that might prove effective in MCDS. We were immediately attracted to the possibility that the chemical chaperone sodium phenylbutyrate (SPB), which was already used clinically to assist in ammonia excretion where there are defects in the urea cycle, may be effective. SPB had already been shown to effectively suppress ER stress and restore glucose homeostasis in mouse models of obesity‐related diabetes.45 These initial attempts to develop a therapy for MCDS proved an expensive lesson in how not to perform such research. In my enthusiasm and conviction that this approach would work, I spent a large amount of money having rodent pellet diet manufactured containing SPB rather than first conducting cell culture experiments which would have been quicker and less costly. Mice were fed the diet for several months, and ultimately, it had no effect on growth or growth plate phenotype in our MCDS mice or in mice with MED caused by a matrilin‐3 mutation.46 This is possibly because SPB exerts its effects at relatively high concentrations that cannot be achieved in the avascular growth plate. Nevertheless, treatment with SPB has subsequently been shown to be of therapeutic value in several mouse models of related ER stress–associated disorders including amelogenesis imperfecta,47 adult intracerebral haemorrhage caused by Col4a1 mutations,48 and in a model of hyperostosis.49

Work in my own laboratory refocused on developing a cell culture model of MCDS with which we could screen chemicals and drugs that may reduce the ER stress associated with MCDS and therefore be of potential therapeutic value. Ewa Mularczyk developed a transient expression system in HeLa cells that gave high levels of ER stress and was suitable for conducting such a drug screen. Accordingly, she tested many of the ER stress–reducing drugs and chemicals already described in the literature and found that only one drug tested, carbamazepine (CBZ), was capable of preventing the increased ER stress caused by the expression of the COL10A1 p.N617K as measured by increases in BIP, CHOP and sXBP1 mRNA levels.50

CBZ is a drug that has been in clinical use for over 50 years for the control of seizures and other related disorders and has been used effectively in children. It is a known stimulator of autophagy. It has been used in mouse models to prevent a genetic form of hepatic fibrosis induced by the Z mutant form of alpha‐1 antitrypsin, and its effects were shown to prevent the intracellular accumulation of mutant protein by stimulating both autophagic and proteasomal degradation.51 CBZ is currently in clinical trial for the treatment of severe liver disease associated with alpha‐1 antitrypsin deficiency.52

We extended our cell culture studies and demonstrated that CBZ suppressed the ER stress induced by several different MCDS‐causing mutations (NC1del13, G618V and Y598D as well as N617K). The drug worked by stimulating intracellular proteolysis of the mutant ER‐retained protein. Interestingly, for N617K and G618V, the major effect was through the stimulation of proteasomal degradation, whereas NC1del13 and Y598D were degraded through CBZ‐enhanced autophagy.53 These results were most encouraging in that they suggest CBZ will be effective against most if not all MCDS‐causing mutations whatever the preferred route of intracellular degradation. CBZ was next tested in vivo in our original N617K MCDS mouse model where it effectively reduced biochemical evidence of growth plate ER stress, reduced the histopathological expansion of the growth plate hypertrophic zone (Figure 1A), significantly increased vascular invasion (VI) as measured by osteoclast recruitment to the VI front, significantly increased differentiation and terminal height achieved by hypertrophic chondrocytes, and last but by no means least, significantly increased the rate of bone growth in MCDS (but not in wild‐type) mice (Figure 1B).53 We have recently generated a second knock‐in MCDS mouse model (Y632X), which has a significantly more pronounced phenotype than that of the N617K mouse. Nevertheless, CBZ effectively reduced the clinical phenotype significantly increasing bone growth rates and most impressively, completely ameliorating the hip distortion caused by MCDS in mice within 2 weeks of starting treatment.54 These final papers on the effects of CBZ on MCDS in mouse models mark my retirement and the end of my active research in this area.

Figure 1.

Effects of CBZ on growth plate histology and bone growth in the N617K MCDS mouse. A, Three‐week‐old MCDS mice were given CBZ and their growth plates examined after 1 week of treatment (A). Note the expansion of the hypertrophic zone (brackets) seen in the MCDS versus WT mouse was considerably reduced in the MCDS mouse treated for 1 week with CBZ. B, Long bone growth was significantly increased by CBZ treatment in MCDS mice over a 4‐week period (#p>0.05; n=8 per group) [Colour figure can be viewed at wileyonlinelibrary.com]

5. TRANSLATION TO THE CLINIC (2016 ONWARDS)

I am happy to say that the research goes on through the efforts of my collaborator of 20 years, Mike Briggs who moved to the Institute of Human Genetics Newcastle University about 5 years ago, and his clinical collaborator Mike Wright. The potential use of CBZ to treat a number of disorders related to MCDS was patented by Newcastle University (WO2017158356A1) as a first step to developing a clinical trial. In late 2016, CBZ was licensed by the European Medicine Agency for the treatment of MCDS in humans.55 This was followed by a successful application to the EU for a grant (H2020 grant no. 754825) to run a multi‐centre phase II/ III clinical trial of CBZ for the treatment of MCDS.56 The trial, which will involve treating around 40 children with MCDS in six centres, is now recruiting and will report its findings in the next 3‐4 years. We all await the outcome with great interest, apprehension and most of all, hope!

6. NEW THERAPIES IN THE PIPELINE

Other than CBZ, more fundamental work on the role of each of the three major pathways controlling ER stress (IRE1, ATF6 and PERK) is generating further possible therapeutics that may be of use for the treatment of MCDS and other related diseases. IRE1 through Xbp1 splicing does not seem to have a significant role in modulating disease severity in MCDS,57 whereas ATF6α activation is an important factor in suppressing disease symptoms in vivo.58 Importantly, recent research has highlighted a drug capable of stimulating ATF6α activation and reducing the consequences of increased ER stress in a number of situations59 and this may prove useful for MCDS. Furthermore, the key role of PERK activation and the resulting increases in ATF4 activity have been shown to be of prime importance in controlling the MCDS phenotype and an inhibitor of this pathway has similar effects to those of CBZ.60 Development of the ATF6α stimulating and PERK inhibiting compounds for clinical use may add significantly to the therapeutic opportunities to effectively treat disorders such as MCDS.

CONFLICTS OF INTEREST

I have no conflicts of interest to declare.

Boot‐Handford RP. Gene cloning to clinical trials—the trials and tribulations of a life with collagen. Int. J. Exp. Path. 2019;100:4–11. 10.1111/iep.12311